AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Why Attention? Analyze BiLSTM Deficiency and Its Remedies in the Case of NER

Peng-Hsuan Li, Tsu-Jui Fu, Wei-Yun Ma

AAAI 2020 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 68 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Interleave forward and backward LSTM layers to model cross-contextBaseline-BiLSTM-CNN · OntoNotes 5.0 testF1: 87.75±0.14 -> 88.27±0.17Improved
Add token-level self-attention over BiLSTM statesBaseline-BiLSTM-CNN · OntoNotes 5.0 testF1: 87.75±0.14 -> 88.40±0.18Improved
Interleave forward and backward LSTM layers to model cross-contextBaseline-BiLSTM-CNN · WNUT 2017 testF1: 40.68±1.78 -> 42.85±0.99Improved
Add token-level self-attention over BiLSTM statesBaseline-BiLSTM-CNN · WNUT 2017 testF1: 40.68±1.78 -> 42.26±0.82Improved
Interleave forward and backward LSTM layersCross-BiLSTM-CNN · WNUT 2017 testF1: The paper reports 2.2% improvement over baselineImproved
Use cross-context modeling for three-token entity mentionsCross-BiLSTM-CNN · WNUT 2017, mention length 3F1 improvement vs. Baseline: +8.7%Improved
Use self-attention for three-token entity mentionsAtt-BiLSTM-CNN · OntoNotes 5.0, mention length 3F1 improvement vs. Baseline: +2.3%Improved
Zero token hidden states H at test time (ablation)Att-BiLSTM-CNN · OntoNotes 5.0 dev, I tagper-token recall: -28.18Hurt
Keep only context vectors C all at test time (ablation)Att-BiLSTM-CNN · OntoNotes 5.0 dev, I tagper-token recall: -3.80Hurt
Keep only attention head C2 at test time (ablation)Att-BiLSTM-CNN · OntoNotes 5.0 dev, E tagper-token recall: -36.45Hurt
Keep only attention head C3 at test time (ablation)Att-BiLSTM-CNN · OntoNotes 5.0 dev, E tagper-token recall: -39.19Hurt
Keep only attention head C4 at test time (ablation)Att-BiLSTM-CNN · OntoNotes 5.0 dev, B tagper-token recall: -12.21Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy